一种基于改进YOLOv5s网络的结直肠腺瘤实时检测方法

doi:10.3969/j.issn.1000-1565.2022.03.016

摘要/Abstract

摘要： 为解决现有的基于深度学习的结直肠息肉检测算法计算复杂或检测精度较低,不能在速度和精度方面同时满足实时检测的问题,提出了一种基于单阶段目标检测网络YOLO(you only look once)v5s的结直肠腺瘤实时检测方法.在YOLOv5s的主干网中融入通道注意力机制,并以BCEWithLogitsLoss代替其原有的交叉熵损失函数BCELogits,以此提升网络性能.选取2 074张腺瘤图片和包含19 700帧的20段腺瘤视频,按照3∶1的比例构建结直肠腺瘤训练集和测试集.测试结果表明,结直肠腺瘤检测的平均精度为93.6%,检测速度为93帧/s,验证了该系统可以在肠镜的退镜过程中实时检测腺瘤性息肉,且具有较好的检测性能.

Abstract: There are two major issues in the existing deep learning-based algorithms for colorectal polyp detection which are leading to the failure of real-time detection: one is the complicate calculation and the other is poor detection precision. In this paper, we proposed a real-time method for colorectal adenoma detection based on a single-stage target detection network YOLOv5s, in which the channel attention mechanism was added to the backbone and the loss function of BCEWithLogitsLoss was adopted instead of the original one of BCELogits. We collected 2074 endoscopic adenoma images and 20 adenoma videos with 19700 frames, and then assigned them as training dataset and testing dataset according to the ratio of 3∶1. - DOI:10.3969/j.issn.1000-1565.2022.03.016一种基于改进YOLOv5s网络的结直肠腺瘤实时检测方法刘爽¹,田兆星¹,李浩然¹,常颖²,吴思蓓¹,薛林雁¹(1.河北大学质量技术监督学院,计量仪器与系统国家地方联合工程研究中心,河北保定 071002;2.河北大学附属医院消化内科,河北保定 071000)摘要:为解决现有的基于深度学习的结直肠息肉检测算法计算复杂或检测精度较低,不能在速度和精度方面同时满足实时检测的问题,提出了一种基于单阶段目标检测网络YOLO(you only look once)v5s的结直肠腺瘤实时检测方法.在YOLOv5s的主干网中融入通道注意力机制,并以BCEWithLogitsLoss代替其原有的交叉熵损失函数BCELogits,以此提升网络性能.选取2 074张腺瘤图片和包含19 700帧的20段腺瘤视频,按照3∶1的比例构建结直肠腺瘤训练集和测试集.测试结果表明,结直肠腺瘤检测的平均精度为93.6%,检测速度为93帧/s,验证了该系统可以在肠镜的退镜过程中实时检测腺瘤性息肉,且具有较好的检测性能.关键词:计算机辅助诊断;腺瘤;卷积神经网络;YOLOv5s;实时检测中图分类号:TP391.7 文献标志码:A 文章编号:1000-1565(2022)03-0327-10A real-time method for colorectal adenoma detection based on an improved YOLOv5s networkLIU Shuang¹, TIAN Zhaoxing¹, LI Haoran¹, CHANG Ying², WU Sibei¹, XUE Linyan¹(1. National & Local Joint Engineering Research Center of Metrology Instrument and System, College of Quality and Technical Supervision, Hebei University, Baoding 071002, China; 2. Department of Gastroenterology, Affiliated Hospital of Hebei University, Baoding 071000, China)Abstract: There are two major issues in the existing deep learning-based algorithms for colorectal polyp detection which are leading to the failure of real-time detection: one is the complicate calculation and the other is poor detection precision. In this paper, we proposed a real-time method for colorectal adenoma detection based on a single-stage target detection network YOLOv5s, in which the channel attention mechanism was added to the backbone and the loss function of BCEWithLogitsLoss was adopted instead of the original one of BCELogits. We collected 2074 endoscopic adenoma images and 20 adenoma videos with 19700 frames, and then assigned them as training dataset and testing dataset according to the ratio of 3∶1. - 收稿日期:2021-10-14 基金项目:河北省自然科学基金资助项目(H2019201378);河北大学自然科学多学科交叉研究计划资助项目(DXK201914);河北大学校长科研基金资助项目(XZJJ201914;XZJJ201917;XZJJ201918);大中学生科技创新能力培育专项项目(22E50041D) 第一作者:刘爽(1981—),女,河北保定人,河北大学副教授,博士,主要从事生物医学图像处理方向研究.E-mail:lianlianfushi@126.com 通信作者:薛林雁(1981—),女,河北广平人,河北大学副教授,博士,主要从事生物医学图像处理、计算机视觉方向研究.E-mail:lineysnow@163.com第3期刘爽等:一种基于改进YOLOv5s网络的结直肠腺瘤实时检测方法The experimental results show that the improved network achieved a mean average precision of 93.6% and a detection speed of 93FPS, which implies that the system can detect adenomatous polyps in real time during the retraction of colonoscopy, and outperforms the original state-of-the-art method.

刘爽,田兆星,李浩然,常颖,吴思蓓,薛林雁. 一种基于改进YOLOv5s网络的结直肠腺瘤实时检测方法[J]. 河北大学学报(自然科学版), 2022, 42(3): 327-336.

LIU Shuang, TIAN Zhaoxing, LI Haoran, CHANG Ying, WU Sibei, XUE Linyan. A real-time method for colorectal adenoma detection based on an improved YOLOv5s network[J]. Journal of Hebei University(Natural Science Edition), 2022, 42(3): 327-336.

参考文献

[1] SUNG H, FERLAY J, SIEGEL R L, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA: A Cancer Journal for Clinicians, 2021,71(3):209-249. DOI:10.3322/caac.21660.
[2] 赵胜兵,王树玲,方军,等.国内外结直肠癌早诊早治现状[J].中华消化内镜杂志, 2019(2): 143-147. DOI:10.3760/cma.j.issn.1007-5232.2019.02.016
[3] ZAUBER A G, WINAWER S J, O'BRIEN M J, et al. Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths[J]. N Engl J Med, 2012, 212(8): 408-408. DOI:10.1016/j.rce.2012.04.012.
[4] WINAWER S, FLETCHER R, REX D, et al. Colorectal cancer screening and surveillance: Clinical guidelines and rationale -Update based on new evidence[J]. Gastroenterology, 2003, 124(2): 544-560. DOI:10.1053/gast.2003.50044.
[5] ANDERSON J C, BUTTERLY L F. Colonoscopy: Quality indicators[J]. Clinical and Translational Gastroenterology, 2015, 6(2):77. DOI:10.1038/ctg.2015.5.
[6] MORGAN M B, MATES J L. Applications of artificial intelligence in breast imaging[J]. Radiologic Clinics of North America, 2021, 59(1): 139-148. DOI:10.1016/j.rcl.2020.08.007.
[7] ABDELHAFIZ D, NABAVI S, AMMAR R, et al. Convolutional neural network for automated mass segmentation in mammography[J]. BMC bioinformatics,2020,21(1):1-19. DOI:10.1109/iccabs.2018.8542071.
[8] PEREIRA T, FREITAS C, COSTA J L, et al. Comprehensive perspective for lung cancer characterisation based on I solutions using CT images[J]. Journal of Clinical Medicine, 2020, 10(1): 118. DOI:10.3390/jcm10010118.
[9] CHO J, KIM J, LEE K J, et al. Incidence lung cancer after a ngative CT screening in the national lung screening trial: Deep learning-based detection of missed lung cancers[J]. Journal of Clinical Medicine, 2020, 9(12): 3908. DOI:10.3390/jcm9123908.
[10] TONOZUKA R, MUKAI S, ITOI T. The role of artificial intelligence in endoscopic ultrasound for pancreatic disorders[J]. Diagnostics, 2020, 11(1): 18. DOI:10.3390/diagnostics11010018.
[11] KARKANIS S A, IAKOVIDIS D K, MAROULIS D E, et al. Computer-aided tumor detection in endoscopic video using color wavelet features[J]. IEEE Transactions on Information Technology in Biomedicine, 2003, 7(3): 141-152. DOI:10.1109/TITB.2003.813794.
[12] TIAN Y. Artificial intelligence image recognition methed based on convolutional neural network algorithm[J].IEEE Access,2020,8:125731-125744.DOI:10.1109/ACCESS.2020.3006097.
[13] QADIR H A, BALASINGHAM I, SOLHUSVIK J, et al. Improving automatic polyp detection using cNN by exploiting temporal dependency in colonoscopy video[J]. IEEE Journal of Biomedical and Health Informatics, 2020, 24(1): 180-193. DOI:10.1109/JBHI.2019.2907434.
[14] ZHANG R, ZHENG Y, MAK T W C, et al. Automatic detection and classification of colorectal polyps by transferring Low-Level CNN features from nonmedical domain[J]. IEEE Journal of Biomedical and Health Informatics, 2017, 21(1): 41-47. DOI:10.1109/JBHI.2016.2635662.
[15] MA Y, CHEN X, SUN B. Polyp detection in colonoscopy videos by bootstrapping via temporal consistency[C] //Proceedings -International Symposium on Biomedical Imaging, 2020, 1360-1363. DOI:10.1109/ISBI45749.2020.9098663.
[16] 张文明,姚振飞,高雅昆,等.一种平衡准确性以及高效性的显著性目标检测深度卷积网络模型[J].电子与信息学报, 2020, 42(5): 1201-1208. DOI:10.11999/JEIT190229.
[17] HE K, GKIOXARI G, DOLLAR P, et al. Mask R-CNN[C] //Proceedings of the IEEE International Conference on Computer Vision, 2017: 2980-2988. DOI:10.1109/ICCV.2017.322.
[18] REN S, HE K, GIRSHICK R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149. DOI:10.1109/TPAMI.2016.2577031.
[19] LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single shot multibox detector[C] //European Confrence on Cormputer Vision,Springer,Cham, 2016, 9905 LNCS: 21-37. DOI:10.1007/978-3-319-46448-0_2.
[20] NEUBECK A, VAN GOOL L. Efficient non-maximum suppression[C] //Proceedings International Conference on Pattern Recognition, 2006, 3: 850-855. DOI:10.1109/ICPR.2006.479.
[21] LIU Y, LU B, PENG J, et al. Research on the Use of YOLOv5 object detection algorithm in mask wearing recognition[J]. World Scientific Research Journal, 2020, 6(11): 276-284. DOI: 10.6911/WSRJ.202011_6(11).0038.
[22] ZHANG X, DONGX, WEI Q, et al.Real-time object detection algorithm base on improved YOLO v3[J].Journal of Electronie Imaging, 2019, 28(5): 1-12. DOI:10.1117/1.JEI.28.5.053022.
[23] HU J, SHEN L, ALBANIE S, et al. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8): 2011-2023. DOI:10.1109/TPAMI.2019.2913372.
[24] WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional block attention module[C] //Lecture Notes in Computer Science(including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2018, 11211 LNCS: 3-19. DOI:10.1007/978-3-030-01234-2_1.
[25] 陈勇,刘曦,刘焕淋.基于特征通道和空间联合注意机制的遮挡行人检测方法[J].电子与信息学报, 2020, 42(6): 1486-1493. DOI:10.11999/JEIT190606.
[26] REZATOFIGHI H, TSOI N, GWAK J, et al. Generalized intersection over union: A metric and a loss for bounding box regression[C] //Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2019: 658-666. DOI:10.1109/CVPR.2019.00075.
[27] BA J L KINGMA D P. Adam: A methed for stochastic optimization[C] //3 rd International confrence on learning Representation, 2015:1-15. DOI: 10.48550/arXiv. 1412.6980. (